Segmenting the external surface of a human skull in MRI data by adding shape information to gradient vector flow snakes

This paper presents a method for segmenting bone in magnetic resonance images. The starting point is the formulation of a Gradient Vector Flow (GVF) snake extended to include statistical shape information. This method exploits the diffusion process used by the GVF snake and improves its capacity to deal with occlusion problems by adding a shape term to the traditional scheme. A dataset of coronal MR images of the Visible Human Project is used to extract an outline of the skull region. The results of the segmentation are compared with a segmentation of bone areas taken from CT scans of the same individual. For the MR data, the addition of a shape term improves on the results obtained by using a GVF snake alone.